Podcast · 475 episodes
Best AI papers explained, page 16
by Enoch H. Kang · English
Cut through the noise. We curate and break down the most important AI papers so you don’t have to.
All episodes, page 16
Spectrum tuning: Post-training for distributional coverage and in-context steerabilityHow post-training instruction tuning can damage a model’s conditional distributional modeling (CDM), reducing in-context steerability, output diversity/coverage, and distributional alignment; and how…11 Oct 2025 · 16 min · 10 chapters
Understanding Prompt Tuning and In-Context Learning via Meta-LearningExplains why LLM prompting works using a Bayesian meta-learning view of in-context learning, and compares hard prompt tuning vs prefix tuning (soft prompts), including when prompting fails and how…11 Oct 2025 · 14 min · 6 chapters
MLPs Learn In-Context on Regression and Classification tasksCompares in-context learning (ICL) across MLPs, MLP mixers, and transformers on synthetic regression/classification and relational reasoning tasks, focusing on compute efficiency, context length…11 Oct 2025 · 16 min · 10 chapters
Is Pre-Training Truly Better than Meta-Learning?Few-shot learning comparison between pre-training (PT) + head fine-tuning and model-agnostic meta-learning (MAML), arguing PT isn’t universally better; results depend on task/data diversity measured…11 Oct 2025 · 21 min · 11 chapters
Agentic Context Engineering: Evolving Contexts for Self-Improving Language ModelsAgentic Context Engineering (ACE) for self-improving LLM agents—how to adapt task context over time without catastrophic forgetting or “context collapse,” using structured, incremental updates and…11 Oct 2025 · 18 min · 9 chapters
Do LLMs Recognize Your Preferences? Evaluating Personalized Preference Following in LLMsThe episode discusses research on whether large language models can recognize and consistently follow user preferences over long conversations, and why they often “forget” (the “zero-shot crisis”).9 Oct 2025 · 16 min · 9 chapters
Learning dynamics of LLM finetuningLearning dynamics in LLM fine-tuning—how SFT and DPO change predictions across unrelated examples, causing hallucinations and repetition via an ENTK-driven “squeezing effect.”9 Oct 2025 · 12 min · 7 chapters
Iterative Data Smoothing: Mitigating Reward Overfitting and Overoptimization in RLHFRLHF alignment failures caused by reward overfitting (reward model generalization worsens after ~1 epoch) and reward overoptimization (policy exploits a flawed reward proxy, reducing ground-truth…9 Oct 2025 · 17 min · 7 chapters
OpenAI Agent Builder and n8n: Orchestrating Reasoning Versus Automating ProcessCompares OpenAI Agent Builder (AgentKit) vs n8n for enterprise workflows, framing a “reasoning vs orchestration” split and arguing for a hybrid stack.8 Oct 2025 · 15 min · 9 chapters
Training Agents Inside of Scalable World ModelsGoogle DeepMind’s Dreamer 4 trains agents inside scalable world models using offline “imagination” in Minecraft, solving the diamond-finding challenge as a 20,000+ action, pixel-observation…8 Oct 2025 · 14 min · 10 chapters
Small Language Models are the Future of Agentic AIThe episode argues that agentic AI should shift from relying on centralized large language models (LLMs) to an “SLM-first” architecture using small language models (under ~10B parameters) for most…7 Oct 2025 · 19 min · 9 chapters
Activation Steering in Generative Settings via Contrastive Causal Mediation AnalysisContrastive Causal Mediation Analysis (CCM) for precisely locating and steering internal LLM components to control freeform generative behaviors, using contrastive prompt/response pairs and efficient…6 Oct 2025 · 18 min · 9 chapters
Eliciting Secret Knowledge from Language ModelsThe episode covers “secret elicitation” in language models—how models can be fine-tuned to conceal knowledge they still use, and how researchers audit/expose that concealment using black-box and…6 Oct 2025 · 15 min · 7 chapters
Temporal difference flowLong-horizon reinforcement learning for predictive AI, focusing on temporal difference flows (TD flow) to train geometric horizon models (GHMs) without unstable, noisy learning.6 Oct 2025 · 15 min · 8 chapters
Personalized reasoning: just-in-time personalization and why LLMs fail at itThe episode argues that LLMs often get “correctness” but fail “just-in-time personalization,” because training typically separates factual accuracy from preference alignment.5 Oct 2025 · 14 min · 5 chapters
Prompt Curriculum Learning for Efficient LLM Post-TrainingPrompt Curriculum Learning (PCL) for efficient LLM post-training, aiming to improve reasoning (e.g., math) without the heavy cost of RL with verifiable rewards (RLVR).5 Oct 2025 · 13 min · 6 chapters
Personalizing Reinforcement Learning from Human Feedback with Variational Preference LearningPluralistic alignment for AI systems—personalizing reinforcement learning from human feedback (RLHF) so models don’t average conflicting user preferences.4 Oct 2025 · 18 min · 13 chapters
Enhancing Personalized Multi-Turn Dialogue with Curiosity RewardCuriosity-driven user modeling reward (Cure.io) for personalized multi-turn dialogue, aiming to avoid “one-size-fits-all” recommendations by giving LLMs dense, intrinsic feedback during chats.4 Oct 2025 · 14 min · 7 chapters
Learning to summarize user information for personalized reinforcement learning from human feedbackPLUS (pluralistic alignment via summarization) improves RLHF personalization for diverse users by replacing a single-average Bradley-Terry-Luce reward model with user-specific, text-based preference…4 Oct 2025 · 16 min · 8 chapters
Distributional Preference Learning: Understanding and Accounting for Hidden Context in RLHFHow RLHF preference learning hides “hidden context” from annotators and, via its aggregation math, implicitly implements a social choice rule (Borda count), leading to strategic manipulability and…3 Oct 2025 · 16 min · 8 chapters
LIMI: Less is More for Agency“Age of AI Agency” and the LIME/LiMI research (“Less is More for Intelligent Agency”) arguing that autonomous AI agents can be trained with far fewer examples by using high-quality “trajectories”…1 Oct 2025 · 14 min · 5 chapters
LoRA Without RegretLoRA (low-rank adaptation) for adapting large language models efficiently, focusing on when it matches full fine-tuning (“low regret”), plus operational benefits, compute/FLOPs savings, and behavior…1 Oct 2025 · 22 min · 9 chapters
Actor-Critic without Actor: Critic-Guided Denoising for RLActor-Critic Without Actor (ACA) for reinforcement learning replaces the usual actor network with critic-guided denoising, aiming to remove policy lag and cut model size while retaining multimodal…29 Sep 2025 · 16 min · 11 chapters
DELTA-Code: How Does RL Unlock and Transfer New Programming Algorithms in LLMs?Whether reinforcement learning (RL) can make large language models (LLMs) acquire genuinely new programming algorithms, and whether that skill transfers beyond the training distribution.29 Sep 2025 · 16 min · 6 chapters